Table 2 Final statements on OA phenotypes

1

OA phenotypes are subtypes of OA that share distinct underlying pathobiological and pain mechanisms and their structural and functional consequences.

86

60—80—94—100

2

OA phenotypes can become apparent in differences in risk factors, prognostic factors, nature and extent of symptoms and signs, disease trajectory, and/or responsiveness to particular treatments or treatment in general.

89

70—80—99—100

3

An OA phenotype classification system is likely to consist of input variables that together reflect (the likelihood of) the presence of one or more pathobiological and pain mechanisms.

88

70—80—94—100

4

Classification systems are likely to use one or more measures from either one or more domains (e.g., imaging markers, biochemical markers, and pain) to identify a clinically relevant OA phenotype or phenotypes.

86

60—80—95—100

5

The potentially identified phenotype(s) should differ from others in terms of clinically relevant disease-driving factors and/or outcomes.

87

50—80—99—100

6

Research efforts may initially lead to multiple proposed phenotype classification systems. Eventually, these should be aligned and come together in one.

84

65—72—94—100

7a

Differences in the disease stage may cause different results from OA phenotyping studies between study populations. It is likely that the nature and course of disease stages may differ between patients and phenotypes.

82

50—80—90—100

7b

Disease stage(s) of the study population should always be reported. Reasons to take or not take disease stage into account in the analyses (e.g., to adjust for confounding or look for interaction) should be weighted for every study.

84

50—80—99—100

8a

Some components of pathobiological and pain mechanisms in OA may be similar between different joints such as knee and hip (e.g., synovitis, central pain perception), while others may differ (e.g., menisci, femoral head shape). The decision to extrapolate findings from one joint to another, or not, should be justified.

86

50—80—94—100

8b

Phenotype classification systems can be designed for individual joints or systemically (e.g., for multiple joints in one patient), depending on the pathobiological and pain mechanism that is under study and the goal of the study.

86

70—80—90—100

9

Data-driven approaches for constructing phenotype classification systems are generally preferable over expert opinion-based approaches, as long as they are performed using high-quality data and appropriate statistics, are reproducible, and have clinical validity, relevance, and applicability as judged by experts in the field.

91

70—86—100—100